INSTITUTE OF AERONAUTICAL ENGINEERING

(Autonomous)
Dundigal, Hyderabad - 500 043

COMPUTER SCIENCE AND ENGINEERING

DEFINITION AND TERMINOLGY

Course Title FOUNDATIONS OF MACHINE LEARNING

Course Code ACAC03
Program B.Tech
Semester IV CSE (AI & ML)
Course Type Core
Regulation IARE - UG
20
Theory Practical
Course Structure Lecture Tutorials Credits Laboratory Credits
3 0 3 1 2
Course Coordinator Dr. Shaik Jakeer Hussain, Associate Professor

COURSE OBJECTIVES:
The students will try to learn:

I The underlying mathematical principles from probability, linear algebra and

optimization.
II The knowledge of using machine learning to make predictions in a scientific
computing environment.
III The knowledge on machine learning algorithms correlated with paradigms of
supervised and un supervised learning.
IV The knowledge advanced supervised learning applications in machine learning,
such as Neural Networks, Support vector machines etc.

COURSE OUTCOMES:
After successful completion of the course, students should be able to:
CO 1 Understand the need for Machine Learning, various learning tasks, and Understand
statistical learning framework
CO 2 Make use of different supervised learning algorithms to solve data Apply
classification problems.
CO 3 Apply the Ensemble and Probabilistic learning techniques to combine the Apply
predictions from two or more models.
CO 4 Acquire the knowledge about different unsupervised learning algorithms Apply
for clustering of the data.
CO 5 Discuss the advanced supervised learning techniques to solve the Apply
classification problems.
CO 6 Apply the algorithms to a real problem, optimize the models learned, Apply
and evaluate their performance efficiency

DEFINITION AND TERMINOLOGY:

S.No DEFINITIONS CO’s
MODULE I
INTRODUCTION TO MACHINE LEARNING
1 Define Machine Learning ?
CO 1
Machine learning algorithms build a model based on sample data,
known as training data, in order to make predictions or decisions
without being explicitly programmed. machine learning algorithms
are used in a wide variety of applications, such as email filtering and
computer vision, where it is difficult or unfeasible to develop
conventional algorithms to perform the needed tasks.
2 Define Classification?
CO 1
Classification is to assign a category to each item. Applications
include document classification, text classification, image
classification where number of categories are small. Other
applications where there are large or unbounded categories are
optical character recognition and speech recognition.
3 What are Features?
CO 1
It is a set of data attributes often represented as a vector associated
to an example. Feature extraction from examples is domain specific
task done by the experts, and is critical to the successful prediction.
If an example has N attributes and all of them can be represented by
real numbers, then the feature set X is equal to R .
4 What are Labels?
CO 1
Values of categories assigned to examples. Labels are discrete for
classification and real valued for regression. The set of labels is
typically denoted by Y
5 What is a Sample? What are different kinds of Samples?
CO 1
Sample is a collection of examples together with their labels is called
a sample. There are three kinds of samples.Training sample,
examples or samples to train a learning algorithm. Validation
sample, samples to tune the free parameters of the learning
algorithm. Test sample, samples to evaluate the performance of the
learning algorithm.
6 What is a Loss Function?
CO 1

Page 2
Loss function is used to measures the difference or loss between
predicted and the true label. Set of predictions are denoted by Y not
necessarily equal to the set of labels Y.
7 What is a Hypothesis set?
CO 1
Hypothesis set is the set H is a subset of Y, X of functions mapping
features to the set of labels.
8 What are the Learning stages of a given sample?
CO 1
Randomly partition into training, validation, and test sample.
Associate features to examples. Fix free learning parameters and
pick a hypothesis. Pick the hypothesis with best performance on
validation sample. Predict labels of the test examples Evaluated the
algorithm using the test labels.
9 What is Supervised Learning
CO 1
The learner receives a sample for training and validation, and makes
prediction for all unseen points. This is common scenarios for
classification, regression, and ranking.
10 Define PAC Learning
CO 1
In computational learning theory, probably approximately correct (
PAC) learning is a framework for mathematical analysis of machine
learning. In this framework, the learner receives samples and must
select a generalization function called the hypothesis.
11 What is Unsupervised Learning?
CO 1
The learner receives unlabeled examples for training and makes
predictions for all unseen points. Difficult to quantitatively evaluate
the performance of a learner. Clustering and dimensionality
reduction are examples of unsupervised learning
12 What is Active Learning?
CO 1
The learner adaptively or interactively collects training samples by
querying an oracle for new samples. The goal is to achieve
comparable performance to the supervised learning with fewer
samples.
13 What is Re-inforced Learning?
CO 1
The learner actively interacts with the environment and receives an
immediate reward for each action. The objective is to maximize
reward over a course of actions and iterations with the environment.

Page 3
14 What is Transductive Inference?
CO 1
The learner receives a labeled training sample along with a set of
unlabeled test points, and make predictions for only these test points
15 What is Online Learning?
CO 1
At each round, the learner receives an unlabeled training example,
makes a prediction, receives the true label, and incurs a loss. The
objective is to minimize the cumulative loss over all rounds.
16 What is a Support Vector Machine?
CO 1
Support Vector Machines are a set of related Supervised Learning
Methods used for Classification and Regression. Given a set of
training examples, each marked as belonging to one of two categories,
an SVM training algorithm builds a model that predicts whether a
new example falls into one category or the other
17 What is Cluster Analysis?
CO 1
Cluster Analysis is the assignment of a set of observations into
subsets called clusters so that observations within the same cluster
are similar according to some predesignated criterion or criteria,
while observations drawn from different clusters are dissimilar.
18 What is a Genetic Algorithm?
CO 1
A genetic algorithm is a search heuristic that mimics the process of
natural selection, and uses methods such as mutation and crossover
to generate new genotype in the hope of finding good solutions to a
given problem.
19 What is Artificial Intelligence?
CO 1
Artificial Intelligence is the intelligence exhibited by machines or
software. It is also the name of the academic field of study which
studies how to create computers and computer software that are
capable of intelligent behavior. It is also defined as the study and
design of intelligent agents.
20 What is Entropy?
CO 1
The entropy, H , of a discrete random variable X intuitively is a
measure of the amount of uncertainty associated with the value of X
when only its distribution is known
MODULE II
SUPERVISED LEARNING ALGORITHMS
1 What is Exploratory Data Analysis?
CO 2
Exploratory data analysis is an approach to analyzing data sets to
summarize their main characteristics, often with visual methods. A
statistical model can be used or not, but primarily EDA is for seeing
what the data can tell us beyond the formal modeling or hypothesis
testing task .

Page 4
2 What is Computational Science?
CO 1
Computational science also scientific computing or scientific
computation is concerned with constructing mathematical models
and quantitative analysis techniques and using computers to analyze
and solve scientific problems.
3 What is Binary Entropy Function?
CO 2
The entropy for a random variable with two outcomes is the binary
entropy function, usually taken to the logarithmic base 2, thus have
the shannon as unit
4 What is Predictive Analysis?
CO 3
Predictive analytics is a variety of statistical techniques from
modeling, machine learning, and data mining that analyze current
and historical facts to make predictions about future, or otherwise
unknown, events.In business, predictive models exploit patterns
found in historical and transactional data to identify risks and
opportunities. Models capture relationships among many factors to
allow assessment of risk or potential associated with a particular set
of conditions, guiding decision making for candidate transactions.
5 What is a Time Series Model?
CO 2
Time series models are used for predicting or forecasting the future
behavior of variables. These models account for the fact that data
points taken over time may have an internal structure (such as
autocorrelation, trend or seasonal variation) that should be
accounted for
6 What is a Classification and Regression Tree?
CO 2
Classification and regression trees are a non-parametric decision tree
learning technique that produces either classification or regression
trees, depending on whether the dependent variable is categorical or
numeric, respectively.
7 What is a Knot?
CO 3
Knot is where one local regression model gives way to another and
thus is the point of intersection between two splines. In multivariate
and adaptive regression splines, basis functions are the tool used for
generalizing the search for knots
8 What is Linear Modelling in Classification?
CO 2
Linear modelling in a classification context consists of regression
followed by a transformation to return a categorical output and
thereby producing a decision boundary. Really there isn’t much to
the model which makes diagnosis somewhat simple.

Page 5
9 Define Multi-class Classification
CO 3
Multi-class classification is the classification technique that allows us
to categorize the test data into multiple class labels present in
trained data as a model prediction.
10 Define Multi-Label Classification
CO 2
Multilabel classification is used when there are two or more classes
and the data we want to classify may belong to none of the classes or
all of them at the same time, e.g. to classify which traffic signs are
contained on an image.
11 Define hypothesis
CO 2
Hypothesis in Machine Learning is used when in a Supervised
Machine Learning, we need to find the function that best maps input
to output. This can also be called function approximation because
we are approximating a target function that best maps feature to the
target.
12 Define Generalization
CO 2
Generalization is a term used to describe a model’s ability to react to
new data. That is, after being trained on a training set, a model can
digest new data and make accurate predictions. A model’s ability to
generalize is central to the success of a model.
13 Define Most Specific Hypothesis
CO 3
A hypothesis, h, is a most specific hypothesis if it covers none of the
negative examples and there is no other hypothesis h that covers no
negative examples, such that h is strictly more general than h.
14 What is Ada Boosting?
CO 2
AdaBoost is the first stepping stone in the world of Boosting.
AdaBoost is one of the first boosting algorithms to be adapted in
solving practices. Adaboost helps you combine multiple weak
classifiers into a single strong classifier.
15 What is Bagging ?
CO 2
Bootstrap Aggregating, also knows as bagging, is a machine learning
ensemble meta-algorithm designed to improve the stability and
accuracy of machine learning algorithms used in statistical
classification and regression. It decreases the variance and helps to
avoid overfitting. It is usually applied to decision tree methods.
16 Define Occam’s razor?
CO 2
A scientific and philosophical rule that entities should not be
multiplied unnecessarily which is interpreted as requiring that the
simplest of competing theories be preferred to the more complex or
that explanations of unknown phenomena be sought first in terms of
known quantities

Page 6
17 Define Regression
CO 2
Simple Linear Regression is a type of Regression algorithms that
models the relationship between a dependent variable and a single
independent variable. The relationship shown by a Simple Linear
Regression model is linear or a sloped straight line, hence it is called
Simple Linear Regression.
18 Define Interpolation
CO 2
Interpolation is a method of deriving a simple function from the given
discrete data set such that the function passes through the provided
data points. This helps to determine the data points in between the
given data ones. This method is always needed to compute the value
of a function for an intermediate value of the independent function.
19 Define Extrapolation
CO 3
In terms of machine learning, extrapolation can be thought of as
being trained on a certain range of data and being able to predict on
a different range of data. This may be easy with simple patterns,
such as simple positive number or negative number
20 Define ill-posed problem
CO 3
A problem which may have more than one solution, or in which the
solutions depend discontinuously upon the initial data
MODULE III
ENSEMBLE AND PROBABILISTIC LEARNING
1 Define Inductive Bias
CO 3
The inductive bias also known as learning bias of a learning
algorithm is the set of assumptions that the learner uses to predict
outputs given inputs that it has not encountered. In machine
learning, one aims to construct algorithms that are able to learn to
predict a certain target output.
2 Define Model Selection
CO 3
Model selection is the task of selecting a statistical model from a set
of candidate models, given data. In the simplest cases, a pre existing
set of data is considered. However, the task can also involve the
design of experiments such that the data collected is well-suited to
the problem of model selection.
3 Define Underfitting
CO 3
Underfitting is a statistical model or a machine learning algorithm is
said to have underfitting when it cannot capture the underlying
trend of the data. It’s just like trying to fit undersized pants,
Underfitting destroys the accuracy of our machine learning model.

Page 7
4 Define Density Estimation?
CO 3
Density Estimation is a structure to the input space such that
certain patterns occur more often than others.
5 Define Overfitting
CO 3
Overfitting is a modelling error which occurs when a function is too
closely fit to a limited set of data points. Overfitting the model
generally takes the form of making an overly complex model to
explain idiosyncrasies in the data under study. In reality, the data
often studied has some degree of error or random noise within it.
6 Define Validation set
CO 3
In machine learning, a validation set is used to tune the parameters
of a classifier. The validation test evaluates the program’s capability
according to the variation of parameters to see how it might function
in successive testing. The validation set is also known as a validation
data set, development set or dev set.
7 Define Cross-Validation
CO 3
Cross validation is a technique that is used for the assessment of how
the results of statistical analysis generalize to an independent data
set. Cross validation is largely used in settings where the target is
prediction and it is necessary to estimate the accuracy of the
performance of a predictive model
8 Define Unobservable variables
CO 3
Observed variables are variables for which you have measurements in
your dataset, whereas unobserved or latent variables are variables for
which you don’t. When your analysis reveals correlations between
observed variables, you might look for unobserved variables to
explain the correlation, especially in cases where you doubt that there
is a direct causal relationship between them. To offer a contrived
example, suppose your dataset includes strongly correlated variables.
9 Define Bayes’ rule
CO 3
Bayes’ rule allows us to compute the single term P (B or A) in terms
of P (A or B), P (B), and P (A). This is very useful in cases where
we have a good probability of these three terms and want to
determine the fourth one.
10 Define Prior probability
CO 3
Prior probability, in Bayesian statistical inference, is the probability
of an event before new data is collected
11 Define Class Likelihood CO 3
A Likelihood function, on the other hand, takes the data set as a
given, and represents the likeliness of different parameters for your
distribution. X is equal to X 1 ,X 2 and so on is f of x or Theta,
where Theta is a parameter. X is equal to x is an observed sample
point. Then the function of Theta defined as is your likelihood
function.

Page 8
12 Define Expected Risk
CO 3
The risk after considering agreed actions that have not yet been
implemented Targeted Risk ,The desired optimal level of risk
Existing Controls, Controls currently existing in the business Actions
Agreed actions to further treat risk
13 Define Discriminant Functions
CO 3
A discriminant is a function that takes an input vector x and assigns
it to one of K classes, denoted by Ck. Discriminating between two
classes is easy. We assign a data point to class C1 if y (x) is greater
than or equal to 0 else to class C2.
14 Define Decision Regions
CO 3
The decision regions are separated by surfaces called the decision
boundaries. These separating surfaces represent points where there
are ties between two or more categories. For a minimum distance
classifier, the decision boundaries are the points that are equally
distant from two or more of the templates.
15 Define Utility Theory
CO 3
The main idea of utility theory is really simple: an agent’s
preferences over possible outcomes can be captured by a function
that maps these outcomes to a real number; the higher the number
the more that agent likes that outcome. The function is called a
utility function.
16 Define Association Rules CO 3
Association rules are created by thoroughly analysing data and
looking for frequent if or then patterns. Then, depending on the
following two parameters, the important relationships are observed,
Support indicates how frequently the if or then relationship appears
in the database.
17 Define Basket Analysis
CO 3
Market Basket Analysis is an example of an analytics technique
employed by retailers to understand customer purchase behaviors. It
is used to determine what items are frequently bought together or
placed in the same basket by customers.
18 Define Apriori Algorithm
CO 3
The Apriori algorithm is an algorithm that attempts to operate on
database records, particularly transactional records, or records
including certain numbers of fields or items.
19 Define Baye’s Estimator
CO 3
In estimation theory and decision theory a Bayes estimator or a
Bayes action is an estimator or decision rule that minimizes the
posterior expected value of a loss function .

Page 9
20 Define Posterior Density
CO 3
In machine learning, Maximum a Posteriori optimization provides a
Bayesian probability framework for fitting model parameters to
training data and an alternative and sibling to the perhaps more
common Maximum Likelihood Estimation framework. Maximum a
posteriori learning selects a single most likely hypothesis given the
data.
MODULE IV
UNSUPERVISED LEARNING
1 Define Relative Square Error
CO 4
The relative squared error is relative to what it would have been if a
simple predictor had been used. More specifically, this simple
predictor is just the average of the actual values. Thus, the relative
squared error takes the total squared error and normalizes it by
dividing by the total squared error of the simple predictor.
2 Define Bias or Variance
CO 4
The Bias variance dilemma or Bias Variance problem is the conflict
in trying to simultaneously minimize these two sources of error that
prevent supervised learning algorithms from generalizing beyond
their training set,The Bias error is an error from erroneous
assumptions in the learning algorithm.
3 Define Multi-Variate Data
CO 4
Multi-Variate data is the data in which analysis are based on more
than two variables per observation. Usually, multivariate data is used
for explanatory purposes.
4 Define Mean Vector
CO 4
The mean vector consists of the means of each variable and the
variance and covariance matrix consists of the variances of the
variables along the main diagonal and the covariances between each
pair of variables in the other matrix positions
5 Define Co-Variance matrix
CO 4
In statistics and probability theory, a square matrix provides the
covariance between each pair of components or elements of a given
random vector is called a covariance matrix. Any covariance matrix
is symmetric and positive semidefinite.
6 Define Sample Mean
CO 4
In sample, refers to the data that you have, and out of sample to the
data you don’t have but want to forecast or estimate. The data
points used to build the model constitute in sample data whereas all
the new data points not belonging to the training sample constitute
out of sample data.

Page 10
7 Define Imputation
CO 4
It is a technique used for replacing the missing data with some
substitute value to retain most of the data or information of the
dataset.
8 Define Multi-variate Analysis
Multivariate analysis is a Statistical procedure for analysis of data
involving more than one type of measurement or observation. It may
also mean solving problems where more than one dependent variable
is analyzed simultaneously with other variables.
9 Define Linear Discriminant
CO 4
Linear discriminant functions is a function that is a linear
combination of the components of x ,g of x is equal to sum of wtx
and w 0 where w is the weight vector and w 0 is the bias
10 Define Naive Bayes Classifier
CO 4
A classifier is a machine learning model that is used to discriminate
different objects based on certain features. A Naive Bayes classifier is
a probabilistic machine learning model that is used for classification
task. The crux of the classifier is based on the Bayes theorem.
11 Define Nearest Mean Classifier
CO 4
In machine learning, a nearest mean classifier or nearest prototype
classifier is a classification model that assigns to observations the
label of the class of training samples whose mean is closest to the
observation.
12 Define Template Matching
CO 4
Template Matching techniques compare portions of images against
one another. Sample image may be used to recognize similar objects
in source image. If standard deviation of the template image
compared to the source image is small enough, template matching
may be used.
13 Define Regularized Discriminant Analysis.
CO 4
This operator performs a regularized discriminant analysis, for
nominal labels and numerical attributes. Discriminant analysis is
used to determine which variables discriminate between two or more
naturally occurring groups, it may have a descriptive or a predictive
objective.
14 Define Bayesian Spam Filtering
CO 4
A Bayesian filter is a filter that learns your spam preferences. When
you mark emails as spam, the system will note the characteristics of
the email and look for similar characteristics in incoming email,
filtering anything that fits the formula directly in to spam for you.

Page 11
15 Define Multi-Variate Linear Regression
CO 4
Multivariate linear regression is a natural generalization of the simple
linear regression model is a situation including influence of more than
one independent variable to the dependent variable, again with a
linear relationship mathematically .
16 Define Principal Component Analysis CO 4
The central idea of principal component analysis is to reduce the
dimensionality of a data set consisting of a large number of
interrelated variables, while retaining as much as possible of the
variation present in the data set.
17 Define Factor Analysis
CO 4
It is a theory used in machine learning and related to data mining.
The theory behind factor analytic methods is that the information
gained about the interdependencies between observed variables can
be used later to reduce the set of variables in a dataset.
18 Define Z- Normalization
CO 4
Z score normalization refers to the process of normalizing every value
in a dataset such that the mean of all of the values is 0 and the
standard deviation is 1. We use the following formula to perform a z
score normalization on every value in a dataset, New value is equal to
x minus u divided by sigma
19 Define Tuning Complexity
CO 4
Time and Space Complexity of a Turing Machine For a Turing
machine, the time complexity refers to the measure of the number of
times the tape moves when the machine is initialized for some input
symbols and the space complexity is the number of cells of the tape
written. Time complexity all reasonable functions minus T of n is
equal to O of n log n
20 What is Dimensionality Reduction?
CO 4
By reducing the number of input features, thereby reducing the
number of dimensions in the feature space. Dimensionality reduction
means reducing your feature set’s dimension. Dimensionality
reduction brings many advantages to your machine learning data

Page 12
MODULE V
ADVANCED SUPERVISED LEARNING
1 What is Self Organizing Map ?
CO 5
Self-organizing maps (SOM) are feed-forward networks that use an
unsupervised learning approach through a process called
self-organization. A Kohonen network consists of two layers of
processing units called an input layer and an output layer. There are
no hidden units.
2 Define Isomap ? CO 5
Isomap is a nonlinear dimensionality reduction method. It is one of
several widely used low dimensional embedding methods.Isomap is
used for computing a quasiisometric, lowdimensional embedding of a
set of high dimensional data points
3 Define Locally Linear Embedding CO 5
An eigenvector method for solving the problem of non linear
dimensionality reduction. The dimensionality reduction by LLE
succeeds in identifying the underlying structure of the manifold.
4 What is Gaussian Mixture Model in Machine Learning?
CO 5
In machine learning, this is known as Clustering. In real life, many
datasets can be modeled by Gaussian Distribution (Univariate or
Multivariate). So it is quite natural and intuitive to assume that the
clusters come from different Gaussian Distributions.
5 Define k-Means Clustering
CO 5
K-Means Clustering is an Unsupervised Learning algorithm, which
groups the unlabeled dataset into different clusters. Here K defines
the number of pre-defined clusters that need to be created in the
process, as if K is equal to 2, there will be two clusters, and for K is
equal to 3, there will be three clusters.
6 Define Color Quantization
CO 5
Color Quantization is a method of reducing the number of colors
required to represent an image. For example, converting a
photograph to GIF format requires the number of colors to be
reduced to 256
7 What is Vector Quantization
CO 5
Vector quantization is used for lossy data compression, lossy data
correction, pattern recognition, density estimation and clustering.
Lossy data correction, or prediction, is used to recover data missing
from some dimensions.
8 Define Re-construction Error
CO 5
The general definition of the reconstruction error would be the
distance between the original data point and its projection onto a
lower dimensional subspace

Page 13
9 Define Expectation Maximization algorithm
CO 5
The essence of Expectation Maximization algorithm is to use the
available observed data of the dataset to estimate the missing data
and then using that data to update the values of the parameters.
10 Define Gaussian Mixtures
CO 5
A Gaussian mixture model is a category of probabilistic model which
states that all generated data points are derived from a mixture of a
finite Gaussian distributions that has no known parameters.
11 Define Hierarchical Clustering
CO 5
Hierarchical clustering is separating data into groups based on some
measure of similarity, finding a way to measure how they are alike
and different, and further narrowing down the data.
12 Define Agglomerative Clustering
CO 5
The agglomerative clustering is the most common type of hierarchical
clustering used to group objects in clusters based on their similarity.
The algorithm starts by treating each object as a singleton cluster.
Next, pairs of clusters are successively merged until all clusters have
been merged into one big cluster containing all objects.
13 Define Divisive Clustering CO 5
The Divisive Clustering Algorithm is a top down clustering
approach, initially, all the points in the dataset belong to one cluster
and split is performed recursively as one moves down the hierarchy.
Initially, all points in the dataset belong to one single cluster.
14 Define Single link Clustering
CO 5
Single linkage clustering is one of several methods of hierarchical
clustering. It is based on grouping clusters in bottom up fashion , at
each step combining two clusters that contain the closest pair of
elements not yet belonging to the same cluster as each other.
15 Define Dendrogram.
CO 5
A Dendrogram is a diagram representing a tree, in Hierarchical
Clustering, it illustrates the arrangement of the clusters produced by
the corresponding analysis.
16 Define Multi Dimensonal Scaling?
CO 5
Multi-dimensional scaling is a tool by which to quantify similarity
judgments. Formally, MDS refers to a set of statistical procedures
used for exploratory data analysis and dimension reduction
17 Define Subset Selection?
CO 5
This is a naive approach that essentially tries to find the best model
among 2 p models that are trained on all possible subsets of the p
variables.

Page 14
18 Define Structural Risk Minimization ? CO 5
Structural Risk Minimization is an inductive principle of use in
machine learning. Commonly in machine learning, a generalized
model must be selected from a finite data set, with the problem of
overfitting, the model becomes too strongly tailored to the
particularities of the training set and poorly generalized to new data.
19 Define Least Squares Estimate?
CO 5
The Least Square Method is the process of finding the best-fitting
curve or line of best fit for a set of data points by reducing the sum
of the squares of the offsets residual part of the points from the
curve. During the process of finding the relation between two
variables, the tre
20 Define Bernouli Density ?
CO 5
Bernoulli Naive Bayes is one of the variants of the Naive Bayes
algorithm in machine learning. It is very useful to be used when the
dataset is in a binary distribution where the output label is either
present or absent.

Course Coordinator: HOD, CSE (AI & ML)

Ms. V Jalaja, Assistant Professor

Page 15